A Distributed, yet Symbolic Model for Text-to-speech Processing 1 Modelling Reading Aloud

In this paper, a data-oriented model of text-to-speech processing is described. On the basis of a large text-to-speech corpus, the model automatically gathers a distributed , yet symbolic representation of subword-phoneme association knowledge, representing this knowledge in the form of paths in a decision tree. Paths represent context-sensitive rewrite rules which unambiguously map strings of letters onto single phonemes. The more ambiguous the mapping is, the larger the stored context. The knowledge needed for converting a spelling word to its phonemic transcription is thus represented in a distributed fashion: many diierent paths contribute to the phonemisation of a word, and a single path may contribute to phonemisations of many words. Some intrinsic properties of the data-oriented model are shown to have relations with psycholinguistic concepts such as a language's orthographic depth, and word pronunciation consistency. Within psycholinguistics, various models have been proposed in which reading aloud in skilled readers is modelled. These models vary considerably in their restrictions imposed on the representation and processing of knowledge. All of these models face the problem how to explain that experienced human readers are able to pronounce words with regular or irregular pronunciations, and also to pronounce words they have not encountered before (e.g. nonwords). The ability to pronounce nonwords and regular words can be explained in terms of a model which has knowledge of typical spelling-to-phonology rules; the pronunciation of words with irregular pronunciations, however, implies that the model needs some word-speciic, lexical knowledge. Furthermore, three general desiderata for a psychological model of skilled reading aloud are: (i) the model is able to account for human performance on the level of word naming reaction times and pronunciation errors, (ii) the model is able to build up its body of knowledge by learning from examples, and 1